Jobs / DevOps, SRE, and IT operations
Can agents do DevOps, SRE, and IT operations work?
Run systems and networks, respond to incidents, manage infrastructure, and support users.
- US jobs
- 2,382,350 (1.5% of employment)
- Typical wage
- $94K a year (employment-weighted mean of occupation medians)
- Occupations
- 11 in O*NET 31.0
- Evidence
- 5 direct benchmarks, 4 also relevant, 0 context only
The answer today
Some tasksNo change since 10 Oct 2025.
Rule: Agents complete 20% or more of the tasks (or projects) on a direct benchmark.
The evidence that sets this step
- SREGym by UIUC and University of Toronto: 72.2% of tasks completed · GitHub Copilot (GPT-5.6 Sol, max), 3 Aug 2026
Next step: Most tasks needs 60% or more of the tasks (or projects) on two direct benchmarks from independent institutions.
How much of the work is measured
21% of the work in this family is measured by direct benchmarks. Some of the work is measured (20% to 50%). On 10 Oct 2025: 21%.
Work = the O*NET 31.0 work activities of the family's occupations, weighted by how important each task is and by US employment. A benchmark measures an activity when its tasks make the agent do it and its grading checks it. Method.
| Work activity (O*NET) | Share of the work | Direct benchmark |
|---|---|---|
| Advise others on the design or use of technologies | 8.2% | Not measured |
| Design computer or information systems or applications | 6.8% | Not measured |
| Set up computer systems, networks, or other information systems | 6.3% | Measured |
| Test performance of computer or information systems | 5.2% | Not measured |
| Communicate with others about specifications or project details | 5.0% | Not measured |
| Train others to use equipment or products | 4.5% | Not measured |
| Monitor operation of computer or information technologies | 4.3% | Measured |
| Implement security measures for computer or information systems | 4.2% | Measured |
| Evaluate the characteristics, usefulness, or performance of products or technologies | 3.6% | Not measured |
| Resolve computer problems | 3.6% | Measured |
Our reading
Agents complete some DevOps and SRE tasks. On SREGym, the best agent finds and fixes 72.2% of 90 injected Kubernetes faults end to end, without background noise, with an LLM judge for the diagnosis, in one row added in August 2026. Other direct evidence is lower or old: 56.2% on ITBench-AA, which scores diagnosis only; 59.32% on AIOpsLab and 36.7% on IaC-Eval, both with 2024 or early 2025 models; 25% on ITBench, last updated in May 2025. No direct benchmark measures advising others on the use of technologies or designing information systems, the two largest work activities. A second benchmark from an independent institution at 60% or more would move the family to most tasks.
Evidence: SREGym, ITBench-AA, AIOpsLab, IaC-Eval, ITBench by Artificial Analysis, Cisco Research, IBM Research, IISc, Microsoft Research, UC Berkeley, UIUC, University of Michigan, and University of Toronto. Written 10 Oct 2026.
Benchmarks for this work
Direct evidence
Work benchmarks mapped primary to this family. They set the step when they report how many tasks agents complete.
- 36.7%GPT-4 with retrieval-augmented generation · 2024Emerging
Also relevant
These benchmarks cover part of this family's work, or have no result yet. They do not set its step.
- 45.9%Claude Opus 4.6 · 16 Mar 2026Emerging
- OSWorld-Verifiedby HKU, Salesforce AI Research, Carnegie Mellon University and University of Waterloo90.2%Intelligence-Indeed Agent · 25 Jul 2026Saturated
- 64.8%Claude Code + Opus 5.5 (max) · 6 Oct 2026Strong
Retired
Their maintainers or a major evaluator stopped using them as a frontier measure. They do not count.
- 84.7%NexAU-AHE + GPT-5.5 · 23 Apr 2026Retired
Functions: Incident responseTerminal operationsEnterprise workflowsCustomer serviceInfrastructure as codeCompliance operationsFinOpsComputer useSpreadsheet workFeature developmentML engineering
What we do not know
- Benchmarks measure tasks, not whole jobs. A job is a bundle of tasks, judgment, and relationships.
- Capability is not adoption. A task that agents can do in a benchmark may still be done by people at work.
- Most results come from single runs. Reliability over many runs is often unknown.
Occupations in this family
Sorted by US employment (BLS OEWS May 2025). O*NET sub-occupations share the employment figure of their SOC group. Work measured = the share of the occupation's O*NET work that direct benchmarks test.
| Occupation | SOC | US jobs | Work measured |
|---|---|---|---|
| Computer User Support Specialists | 15-1232.00 | 717K | 21% |
| Computer Systems Analysts | 15-1211.00 | 520K | 14% |
| Health Informatics Specialists | 15-1211.01 | (520K group) | 14% |
| Computer Occupations, All Other | 15-1299.00 | 435K | 11% |
| Computer Systems Engineers/Architects | 15-1299.08 | (435K group) | 11% |
| Web Administrators | 15-1299.01 | (435K group) | 11% |
| Network and Computer Systems Administrators | 15-1244.00 | 314K | 40% |
| Computer Network Architects | 15-1241.00 | 180K | 14% |
| Telecommunications Engineering Specialists | 15-1241.01 | (180K group) | 14% |
| Computer Network Support Specialists | 15-1231.00 | 146K | 44% |
| Database Administrators | 15-1242.00 | 70K | 15% |
How to cite
Credit the original work first: the benchmarks listed on this page, and their institutions.
Then, if you used this page:
Can Agents Work. "Can agents do DevOps, SRE, and IT operations work?." https://canagentswork.com/jobs/devops-sre-it/ (accessed 2026-10-11). CC BY 4.0.@misc{caw-jobs-devops-sre-it,
title = {{Can agents do DevOps, SRE, and IT operations work?}},
author = {{Can Agents Work}},
year = {2026},
howpublished = {\url{https://canagentswork.com/jobs/devops-sre-it/}},
note = {Accessed 2026-10-11. CC BY 4.0}
}